Home > Research > Publications & Outputs > An evolutionary approach to fuzzy rule-based mo...
View graph of relations

An evolutionary approach to fuzzy rule-based model synthesis using indices for rules

Research output: Contribution to journalJournal articlepeer-review

<mark>Journal publication date</mark>1/08/2003
<mark>Journal</mark>Fuzzy Sets and Systems
Issue number3
Number of pages14
Pages (from-to)325-338
Publication StatusPublished
<mark>Original language</mark>English


An approach to fuzzy rule-based model (FRB) synthesis from data based on evolutionary algorithm using indices of the rules is presented in the paper. The resulting models are transparent and existing knowledge could easily be incorporated at the initialisation stage. The main difference between the proposed approach and the previous ones is the treatment of a small part of the complete rule set only, which allows an interpretable resulting model to be achieved and the dimension of chromosome considered in the evolutionary algorithm to be significantly reduced. A specific encoding mechanism is presented considering only the rules, which actually participate in the model. They are represented by their indices and membership functions’ parameters. It allows treatment of the problem of structure and parameter identification with practically meaningful dimensions (tens of fuzzy linguistic terms and fuzzy linguistic variables), while most of the other approaches indirectly suppose a small number of inputs and linguistic terms. As a result, the synthesised fuzzy model is significantly more transparent then other black-box types of models like neural networks, polynomial models and also FRB models considering the complete rule set, since a partial set of fuzzy rules (normally some tens) is easy to be inspected and explained. At the same time, this model is significantly more flexible than first principle models. This approach is applied to modelling of components of heating ventilating and air-conditioning systems and validated with real experimental data. It has potential applications in simulation, control and fault detection and diagnosis. (c) Elsevier

Bibliographic note

The final, definitive version of this article has been published in the Journal, Fuzzy Sets and Systems 137 (3), 2003, © ELSEVIER.